Reproduce “DTOR: Decision Tree Outlier Regressor to explain anomalies” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Riccardo Crupi, Daniele Regoli, Alessandro Damiano Sabatino, Immacolata Marano, Massimiliano Brinis, Luca Albertazzi, Andrea Cirillo, Andrea Claudio Cosentini
Researchers from Intesa Sanpaolo introduced DTOR (Decision Tree Outlier Regressor), a model-agnostic XAI framework generating human-interpretable, rule-based explanations for individual anomaly scores produced by various black-box anomaly detection models. It consistently delivers 100% valid and computationally efficient explanations, critical for financial auditing and risk management.
Key insights
- Rule-based XAI can be effectively adapted for regression tasks by learning to estimate continuous anomaly scores, providing nuanced explanations beyond binary classification.
- Guaranteed explanation validity (100%) is crucial for building trust and ensuring practical utility in high-stakes applications where misinterpretations can have severe consequences.
- Computational efficiency and rule interpretability in the original feature space are paramount for the successful operationalization of XAI in real-world, high-dimensional datasets within regulated industries.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.